PromptLoop exposes a single tool 'mcp_to_promptloop' that acts as a bridge/adapter rather than a domain-specific tool. The tool lacks a proper description (empty string in code), has minimal parameter documentation, and no documented output schema. The tool name does not follow verb_noun conventions (should be 'convert_mcp_to_promptloop' or similar). Input schema is present but sparse. This is a utility/meta-tool, not a production-grade agent tool. The codebase itself is a local LLM wrapper, not an MCP server, it consumes MCP servers but does not expose itself as one.

• Add a substantive description to mcp_to_promptloop: 'Converts a FastMCP server instance into a PromptLoop-compatible tools list and handler callable. Use this to integrate external MCP tools with the local LLM engine. Returns (tools_list, tool_handler_fn) where tools_list is a list of tool schemas and tool_handler_fn dispatches calls by name.'
• Rename tool to 'convert_mcp_server_to_promptloop' or 'adapt_mcp_for_promptloop' to follow verb_noun convention and clarify the action.
• Document the mcp_server parameter: 'A FastMCP server instance with registered tools. Must have list_tools() and call_tool(name, args) methods.'
• Document the return value: 'A tuple (tools, handler) where tools is a list of dicts with keys {name, description, parameters} and handler is a callable(tool_name: str, arguments: dict) -> str that executes the tool and returns text output.'
• Consider whether this library should expose an MCP server interface itself (e.g., via stdio or HTTP) so it can be used as a tool provider by other agents, not just as a library.